Thanks to visit codestin.com
Credit goes to www.scribd.com

Scribd
Upload
27 views7 pages

Week13 Lectures

The document discusses logistic regression analysis, focusing on binary response variables, specifically in the context of coronary heart disease. It explains the Bernoulli distribution of the response variable, the formulation of the regression model, and the likelihood function used for parameter estimation. The parameters are estimated by maximizing the log-likelihood function, which requires iterative procedures and software.

Uploaded by

c7m4n2g5dm
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
27 views7 pages

Week13 Lectures

The document discusses logistic regression analysis, focusing on binary response variables, specifically in the context of coronary heart disease. It explains the Bernoulli distribution of the response variable, the formulation of the regression model, and the likelihood function used for parameter estimation. The parameters are estimated by maximizing the log-likelihood function, which requires iterative procedures and software.

Uploaded by

c7m4n2g5dm
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

STAC67H3: Regression Analysis

Fall, 2014

Instructor: Jabed Tomal

Department of Computer and Mathematical Sciences


University of Toronto Scarborough
Toronto, ON
Canada

December 1, 2014

Jabed Tomal (U of T) Regression Analysis December 1, 2014 1/7


Logistic Regression
Binary Response Variable

1 In a study of coronary heart disease as a function of age, gender,


smoking history, cholesterol level, percent of ideal body weight,
and blood pressure, the response Y is defined to have two
possible outcomes: person developed heart disease, person did
not develop heart disease. The outcome is binary which may be
coded as 1 (developed heart disease) and 0 (did not develop
heart disease).

Jabed Tomal (U of T) Regression Analysis December 1, 2014 2/7


Logistic Regression
Binary Response Variable

1 The response Yi is Bernoulli with probability distribution



πi if y = 1
P(Yi = y ) =
(1 − πi ) if y = 0
2 The expectation of Yi is

E(Yi ) = πi = (?)β0 + β1 Xi

Jabed Tomal (U of T) Regression Analysis December 1, 2014 3/7


Logistic Regression
Binary Response Variable

1 The regression model is obtained by assuming Yi independent


Bernoulli with expected value πi which is a non-linear function of
the explanatory variable X with the following form

exp{β0 + β1 Xi }
πi = .
1 + exp{β0 + β1 Xi }
2 which gives
1
(1 − πi ) = .
1 + exp{β0 + β1 Xi }

Jabed Tomal (U of T) Regression Analysis December 1, 2014 4/7


Logistic Regression
Likelihood Function

1 The probability function for the ith observation Yi is

P(Yi |β0 , β1 ) = πiYi (1 − πi )1−Yi


2 Since the observations are assumed independent, the joint
probability function is obtained as
n n
πiYi (1 − πi )1−Yi
Y Y
P(Y1 , Y2 , · · · , Yn |β0 , β1 ) = P(Yi |β0 , β1 ) =
i=1 i=1

Jabed Tomal (U of T) Regression Analysis December 1, 2014 5/7


Logistic Regression
Likelihood Function

1 The likelihood function is obtained by considering the joint


probability function as a function of β0 and β1 given the
observations Y1 , Y2 , · · · , Yn as following
n
πiYi (1 − πi )1−Yi
Y
L(β0 , β1 |Y1 , Y2 , · · · , Yn ) =
i=1

Jabed Tomal (U of T) Regression Analysis December 1, 2014 6/7


Logistic Regression
Likelihood Function

1 The log likelihood function is obtained by taking log on both sides


of the likelihood function

n
X
log L(β0 , β1 |Y1 , Y2 , · · · , Yn ) = [Yi log πi + (1 − Yi ) log(1 − πi )]
i=1

where πi is a function of β0 and β1 .


2 The estimators of the parameters β0 and β1 are obtained by
maximizing the log-likelihood function with respect to β0 and β1 ,
respectively. The estimation of the parameters require iterative
procedures and a computer software.

Jabed Tomal (U of T) Regression Analysis December 1, 2014 7/7

You might also like